Could we be on the verge of an era where fighter jets take flight without pilots – and are controlled by artificial intelligence (AI)? US R Adm Michael Donnelly recently said that an upcoming combat jet could be the Navy’s last one with a pilot in the cockpit. That marks a striking, if not entirely surprising, shift in thinking about the future of aerial warfare.

The US Navy is not alone. Other programmes to develop next-generation fighter jets are also touting uncrewed options as a distinct possibility.

However, we have been here before. Senior leaders in the US Navy said they believed the last crewed fighter jet had been procured in 2015. As far back as 1957, premature obituaries were being written for the fighter pilot era. So, is there anything different now?

The ability of a fighter jet to manoeuvre, accelerate, and maintain high speeds, crucial for air combat, is called kinematic performance. Estimates are as high as 80% on how much pilots reduce kinematic performance. Though this figure may be disputed, there is no question that uncrewed aircraft enjoy several key advantages.

Without the need for life support systems such as ejection seats and oxygen supplies, these aircraft can perform in ways that are beyond the scope of piloted aircraft. But additional trends are pushing militaries to reconsider the role of the human pilot altogether.

Systems enabled by AI are already demonstrating superior performance in military exercises. In existing remotely piloted aircraft, a human operator remains in control. This model is known as “human-in-the-loop”. AI is now enabling the possibility of human-on-the-loop (where humans take a step back, supervising and intervening if necessary) and even “human-out-of-the-loop” systems (in which AI selects and engages targets autonomously).

The latter category, while controversial, may offer decisive advantages. In scenarios where milliseconds matter, a fully autonomous system could outperform any human operator, to the extent that senior defence leaders have expressed a willingness to trust AI with lethal decision-making under certain conditions. Others add that autonomous systems could adhere more rigorously to the laws of armed conflict compared with a human operator.

Unpiloted combat jets also offer potential financial savings. Fighter jets are expensive to build, operate and maintain, not least because of the training and equipment needed to support pilots. A 2011 study found that the life cycle cost of a surveillance drone was roughly half that of a comparable piloted platform. And cheaper aircraft are important because of the likely losses which will be inflicted on air forces in the event of a conflict with Russia or China.

Another advantage of fully autonomous aircraft is risk mitigation. As Nato militaries grapple with a shortage of trained pilots for potential conflicts between states, uncrewed systems offer a way to restore the balance without putting lives at risk of death or capture.

Therefore, one option for militaries is to expand the use of remotely piloted aircraft – drones similar to those deployed in Iraq and Afghanistan. Crucially, this would ensure humans maintain control over weapons use. The only difference with the present would be in making these systems the backbone of the fleet, rather than supplementary systems struggling to operate in hostile airspace. This would require upgrading them with state-of-the-art technologies like stealth. This helps aircraft reduce their chances of being detected by the enemy’s radar and infrared (heat) sensors.

A step up from this would be autonomous combat aircraft, carrying the advantages of on- or off-the-loop technologies. The US Air Force’s Project Venom is training AI in modified F-16 jets for eventual transfer to drones. These drones will operate alongside crewed aircraft, as part of mixed human and machine teams. But if this AI software was retained on the F-16s (or transferred to more advanced fighter jets), it could produce a squadron of autonomous jets just as capable as those piloted by humans.

A more radical idea is to forgo traditional fighter jets altogether. Proponents of this vision imagine swarms of low-cost, expendable drones working together to overwhelm enemy defences. While current drones have limitations in range, payload, and labour requirements, true “swarming” could change the equation.

Current limitations

So what is stopping militaries from pressing ahead with these options? A few things. AI isn’t ready, yet. Machine learning – a subset of AI where algorithms learn from experience – underpins all this. But it still struggles with the inherent ambiguity and creativity of war. Simply putting tyres on an aircraft can thwart computer vision – the field of AI that allows computers to interpret images and videos. So training AI to operate in the full range of possible combat situations is a mammoth task. In the words of one air force commander, “robotified warfare…is centuries away”.

Another issue concerns communications, since remotely operated drone systems, especially interconnected, swarming ones, need data links. Given how much adversaries are investing in jamming these signals, designs may be pushed in opposite directions: either keeping a pilot onboard or embracing autonomy so the aircraft can keep fighting, even if it is cut off.

Yet the real limit may be a fear of crossing the Rubicon. While the US and its allies have a de facto “no first use” policy on fully autonomous weapons, the demands of warfare against an enemy willing to use such systems may erode these norms.

So, the navy’s statement is a warning: the age of the human fighter pilot might be ending. But it’s the next war that could make that decision for us.